A Statistical Framework to Quantify Biological Rhythms in Genome-Scale Data

Abstract

Circadian rhythms are near 24-hour patterns produced by endogenous molecular oscillators, or clocks to coordinate systemic processes such as metabolism, sleep, and immunity. Studies extensively use genome-scale data with complex time-series designs to investigate circadian genomic rhythms, which may reveal pathways of circadian association to disease as well as mechanisms for systemic coordination between tissue clocks. Quantifying effect sizes of rhythmic parameters such as amplitude and phase are important when analyzing rhythmicity in one condition, changes to rhythmicity between conditions and disease states (differential rhythmicity), and inferring relationships between clocks. Here, we address statistical challenges to rhythm detection at the genome level by developing a statistical approach to quantify rhythmic effect sizes in order to better investigate circadian gene regulation. We further demonstrate the method’s utility in several applications using various publicly available genome-scale data.